A company that doesn't make chips, write code, or operate data centers is quietly becoming one of the year's hottest plays on artificial intelligence infrastructure. Texas Pacific Land (TPL), which traces its roots to an 1880s railroad bankruptcy, owns roughly 880,000 acres of surface land, mostly spread across the Permian Basin of West Texas. The stock has surged about 38% year to date, far outpacing the S&P 500's roughly 7% rise, even after pulling back from a peak gain of about 91% in late February.

Why Is a Land Company Suddenly Valuable to AI?

The answer lies in a simple geographic coincidence. The same West Texas acreage that sits atop oil and gas reserves also offers the three critical resources that AI data center operators desperately need: cheap land, abundant water for cooling systems, and access to power generation. The Permian Basin produces enormous volumes of natural gas, and developers are increasingly building their own gas-fired power plants next to data centers rather than waiting years to connect to the traditional electrical grid.

Texas Pacific Land doesn't drill or operate wells itself. Instead, it owns the ground and collects payments from energy companies that do, a model that produces enormous margins and demands little spending. In 2025, the company generated $798 million in revenue and $481 million in net income while carrying no debt. First-quarter 2026 revenue rose 21% year over year to a record $237 million, with earnings per share climbing to $2.07 from $1.75 a year earlier.

The company has already begun translating this land advantage into data center deals. In June, Texas Pacific Land agreed to provide surface acreage and brackish groundwater to Chevron for Project Kilby, a large power generation facility that will support a customer data center in Reeves County, Texas. Late last year, it invested $50 million in Bolt, an AI infrastructure company chaired by former Google chief executive Eric Schmidt, taking an equity stake and the right to supply water to Bolt's projects. In the first quarter, it agreed to sell a parcel for about $43 million, paid out over roughly 20 years, tied to a separate data center and power project, along with a water supply agreement.

"Virtually every major hyperscaler and AI lab are evaluating large-scale plans in Texas," said Tyler Glover, CEO of Texas Pacific Land.

Tyler Glover, CEO at Texas Pacific Land

What Is Driving the Broader Data Center Power Crisis?

Texas Pacific Land's sudden relevance reflects a much larger infrastructure bottleneck. Goldman Sachs projects that U.S. data center power demand will reach 41 gigawatts in 2026 and 66 gigawatts in 2027. Meanwhile, at least 75 data center projects worth $130 billion were delayed or blocked in the first quarter of 2026 because of electricity bottlenecks alone. The traditional electrical grid simply cannot keep pace with the speed at which AI companies need to build.

This power crunch has sparked a wave of creative solutions. Sunrun, a residential solar installer, announced a landmark partnership with Tesla and Renew Home on June 24, 2026, to build a virtual power plant (VPP) framework. A VPP is software that connects thousands of small energy assets, such as home batteries, solar systems, and smart thermostats, and coordinates them to behave like a single large generator. The deal targets more than 16 gigawatts of flexible electricity to hyperscalers and utilities, roughly enough to power 12 million U.S. homes simultaneously.

The Sunrun partnership represents a fundamental shift in how data centers will source power. Over 300 megawatts sit ready for immediate deployment in Virginia's Data Center Alley, the highest concentration of U.S. data center infrastructure in the country, with that figure expected to grow to at least 500 megawatts by 2030. The companies describe the platform as deployable in "months, not years," with no new transmission lines required, no land to acquire, and no extra water use on the offtaker's side.

How Are Companies Positioning for AI Power Demand?

• Land and Water Assets: Texas Pacific Land leverages its massive acreage and access to brackish groundwater in the Permian Basin to attract data center developers seeking sites with minimal permitting delays and established power infrastructure nearby.
• Distributed Energy Resources: Sunrun's virtual power plant framework pools home batteries, solar systems, and smart thermostats to create flexible capacity that can be deployed to data centers in months rather than years, avoiding traditional grid interconnection queues.
• Fuel Cell Technology: Ceres Power Holdings is pursuing an asset-light, royalty-driven model based on solid oxide fuel cells (SOFCs) and solid oxide electrolyzers (SOECs), with partnerships from manufacturers like Weichai, Delta, and Doosan to scale production and capture high-margin royalties as AI power demand grows.

Ceres Power, a UK-based fuel cell technology company, represents yet another approach to the power bottleneck. The company operates an asset-light, royalty-driven model centered on solid oxide fuel cells and solid oxide electrolyzers, which can generate electricity from natural gas or hydrogen with high efficiency. Ceres maintains partnerships with manufacturers Weichai, Delta, and Doosan that underpin high-margin, volume-based royalties. Despite current unprofitability and execution risks, Ceres maintains a strong cash buffer and expects operating cost reductions to narrow losses as partner production ramps.

What Are the Risks and Limitations of These Plays?

While the opportunity is real, investors should approach these stories with caution. Texas Pacific Land's data center deals, though promising, remain small compared with the company's primary revenue sources, which are mostly oil and gas royalties and water sales tied to drilling. Those royalties move with energy prices, which can be volatile. At about $396 per share, the stock trades at about 54 times earnings, a steep multiple that already builds in years of growth from data centers that have only begun to contribute.

Sunrun's virtual power plant deal, announced on June 24, 2026, is a framework, not a signed contract. The 16-gigawatt figure is an ambition, not committed revenue. No hyperscaler has yet signed a binding offtake agreement, so the recurring income analysts are excited about doesn't exist in any binding form. The stock is still down year to date, and cash generation was negative in the first quarter when excluding safe harbor equipment investments.

Regulatory exposure also looms large. Sunrun's business depends heavily on federal incentive programs, net metering rules, and utility cooperation. Any of those can shift. The AI infrastructure permitting environment is already under community and political pressure in key markets, including Virginia.

The broader lesson is clear: as AI companies race to build data centers, the real bottleneck isn't computing power or chip supply. It's electricity. Companies that control land, water, and power generation capacity in the right locations are suddenly valuable, even if they don't make a single chip. But valuations have already begun to price in years of future growth, making entry points critical for investors considering these plays.